Computer Organization and Assembly Languages Yung-Yu Chuang 2006/09/18 Course overview Computer Organization and Assembly Languages Yung-Yu Chuang 2006/09/18 with slides by Kip Irvine

Logistics Meeting time: 9:10am-12:10pm, Monday Classroom: CSIE Room 102 Instructor: Yung-Yu Chuang Teaching assistants: 謝毓庭/黃子桓 Webpage: http://www.csie.ntu.edu.tw/~cyy/assembly id / password Forum: http://www.cmlab.csie.ntu.edu.tw/~cyy/forum/viewforum.php?f=7 Mailing list: assembly@cmlab.csie.ntu.edu.tw Please subscribe via https://cmlmail.csie.ntu.edu.tw/mailman/listinfo/assembly/

Prerequisites Programming experience with some high-level language such C, C ++,Java …

Textbook Assembly Language for Intel-Based Computers, 5th Edition, Kip Irvine

References Computer Systems: A Programmer's Perspective, Randal E. Bryant and David R. O'Hallaron The Art of Assembly Language, Randy Hyde Michael Abrash' s Graphics Programming Black Book

Grading (subject to change) Assignments (50%) Class participation (5%) Midterm exam (20%) Final project (25%)

Computer Organization and Assembly language It is not only about assembly. I hope to cover Basic concept of computer systems and architecture x86 assembly language

Why taking this course? It is required. It is foundation for computer architecture and compilers. At times, you do need to write assembly code. “I really don’t think that you can write a book for serious computer programmers unless you are able to discuss low-level details.” Donald Knuth

Reasons for not using assembly Development time: it takes much longer to develop in assembly. Harder to debug, no type checking, side effects… Maintainability: unstructured, dirty tricks Portability: platform-dependent

Reasons for using assembly Educational reasons: to understand how CPUs and compilers work. Better understanding to efficiency issues of various constructs. Making compilers, debuggers and other development tools. Hardware drivers and system code Embedded systems Making libraries. Accessing instructions that are not available through high-level languages. Optimizing for speed or space

To sum up It is all about lack of smart compilers Faster code, compiler is not good enough Smaller code , compiler is not good enough, e.g. mobile devices, embedded devices, also Smaller code → better cache performance → faster code Unusual architecture , there isn’t even a compiler or compiler quality is bad, eg GPU, DSP chips, even MMX.

Syllabus (topics we might cover) IA-32 Processor Architecture Assembly Language Fundamentals Data Transfers, Addressing, and Arithmetic Procedures Conditional Processing Integer Arithmetic Advanced Procedures Strings and Arrays Structures and Macros High-Level Language Interface Real Arithmetic (FPU) SIMD Code Optimization

What you will learn Basic principle of computer architecture IA-32 modes and memory management Assembly basics How high-level language is translated to assembly How to communicate with OS Specific components, FPU/MMX Code optimization Interface between assembly to high-level language

Chapter.1 Overview Virtual Machine Concept Data Representation Boolean Operations

Translating Languages English: Display the sum of A times B plus C. C++: cout << (A * B + C); Intel Machine Language: A1 00000000 F7 25 00000004 03 05 00000008 E8 00500000 Assembly Language: mov eax,A mul B add eax,C call WriteInt

Virtual machines Abstractions for computers

High-Level Language Level 5 Application-oriented languages Programs compile into assembly language (Level 4) cout << (A * B + C);

Assembly Language Level 4 Instruction mnemonics that have a one-to-one correspondence to machine language Calls functions written at the operating system level (Level 3) Programs are translated into machine language (Level 2) mov eax, A mul B add eax, C call WriteInt

Operating System Level 3 Provides services Programs translated and run at the instruction set architecture level (Level 2)

Instruction Set Architecture Level 2 Also known as conventional machine language Executed by Level 1 program (microarchitecture, Level 1) A1 00000000 F7 25 00000004 03 05 00000008 E8 00500000

Microarchitecture Level 1 Interprets conventional machine instructions (Level 2) Executed by digital hardware (Level 0)

Digital Logic Level 0 CPU, constructed from digital logic gates System bus Memory

Data representation Computer is a construction of digital circuits with two states: on and off You need to have the ability to translate between different representations to examine the content of the machine Common number systems: binary, octal, decimal and hexadecimal

Binary Representations Electronic Implementation Easy to store with bistable elements Reliably transmitted on noisy and inaccurate wires 0.0V 0.5V 2.8V 3.3V 1

Binary numbers Digits are 1 and 0 (a binary digit is called a bit) 1 = true 0 = false MSB –most significant bit LSB –least significant bit Bit numbering: A bit string could have different interpretations

Unsigned binary integers Each digit (bit) is either 1 or 0 Each bit represents a power of 2: Every binary number is a sum of powers of 2

Translating Binary to Decimal Weighted positional notation shows how to calculate the decimal value of each binary bit: dec = (Dn-1  2n-1) + (Dn-2  2n-2) + ... + (D1  21) + (D0  20) D = binary digit binary 00001001 = decimal 9: (1  23) + (1  20) = 9

Translating Unsigned Decimal to Binary Repeatedly divide the decimal integer by 2. Each remainder is a binary digit in the translated value: 37 = 100101

Binary addition Starting with the LSB, add each pair of digits, include the carry if present.

Integer storage sizes Standard sizes: Practice: What is the largest unsigned integer that may be stored in 20 bits?

Large measurements Kilobyte (KB), 210 bytes Megabyte (MB), 220 bytes Gigabyte (GB), 230 bytes Terabyte (TB), 240 bytes Petabyte Exabyte Zettabyte Yottabyte

Hexadecimal integers All values in memory are stored in binary. Because long binary numbers are hard to read, we use hexadecimal representation.

Translating binary to hexadecimal Each hexadecimal digit corresponds to 4 binary bits. Example: Translate the binary integer 000101101010011110010100 to hexadecimal:

Converting hexadecimal to decimal Multiply each digit by its corresponding power of 16: dec = (D3  163) + (D2  162) + (D1  161) + (D0  160) Hex 1234 equals (1  163) + (2  162) + (3  161) + (4  160), or decimal 4,660. Hex 3BA4 equals (3  163) + (11 * 162) + (10  161) + (4  160), or decimal 15,268.

Powers of 16 Used when calculating hexadecimal values up to 8 digits long:

Converting decimal to hexadecimal decimal 422 = 1A6 hexadecimal

Hexadecimal addition Divide the sum of two digits by the number base (16). The quotient becomes the carry value, and the remainder is the sum digit. 1 1 36 28 28 6A 42 45 58 4B 78 6D 80 B5 Important skill: Programmers frequently add and subtract the addresses of variables and instructions.

Hexadecimal subtraction When a borrow is required from the digit to the left, add 10h to the current digit's value: -1 C6 75 A2 47 24 2E Practice: The address of var1 is 00400020. The address of the next variable after var1 is 0040006A. How many bytes are used by var1?

Signed integers The highest bit indicates the sign. 1 = negative, 0 = positive If the highest digit of a hexadecmal integer is > 7, the value is negative. Examples: 8A, C5, A2, 9D

Two's complement notation Steps: Complement (reverse) each bit Add 1 Note that 00000001 + 11111111 = 00000000

Binary subtraction When subtracting A – B, convert B to its two's complement Add A to (–B) 1 1 0 0 1 1 0 0 – 0 0 1 1 1 1 0 1 1 0 0 1 Advantages for 2’s complement: No two 0’s Sign bit Remove the need for separate circuits for add and sub

Ranges of signed integers The highest bit is reserved for the sign. This limits the range:

Character Character sets Null-terminated String Using the ASCII table Standard ASCII (0 – 127) Extended ASCII (0 – 255) ANSI (0 – 255) Unicode (0 – 65,535) Null-terminated String Array of characters followed by a null byte Using the ASCII table back inside cover of book

IEEE Floating Point IEEE Standard 754 Established in 1985 as uniform standard for floating point arithmetic Before that, many idiosyncratic formats Supported by all major CPUs Driven by Numerical Concerns Nice standards for rounding, overflow, underflow Hard to make go fast Numerical analysts predominated over hardware types in defining standard

Fractional Binary Numbers 4 • • • 2 1 bi bi–1 b2 b1 b0 b–1 b–2 b–3 b–j • • • . 1/2 1/4 • • • 1/8 2–j Representation Bits to right of “binary point” represent fractional powers of 2 Represents rational number:

Binary real numbers Binary real to decimal real Decimal real to binary real 4.5625 = 100.10012

Frac. Binary Number Examples Value Representation 5-3/4 101.112 2-7/8 10.1112 63/64 0.1111112 1/3 0.0101010101[01]…2 1/5 0.001100110011[0011]…2 1/10 0.0001100110011[0011]…2

IEEE floating point format IEEE defines two formats with different precisions: single and double 23.85 = 10111.1101102=1.0111110110x24 e = 127+4=83h 0 100 0001 1 011 1110 1100 1100 1100 1100

IEEE floating point format special values IEEE double precision

Denormalized numbers Number smaller than 1.0x2-126 can’t be presented by a single with normalized form. However, we can represent it with denormalized format. 1.0000..00x2-126 the least “normalized” number 0.1111..11x2-126 the largest “denormalized” numbr 1.001x2-129=0.001001x2-126

Summary of Real Number Encodings  + -Normalized -Denorm +Denorm +Normalized NaN NaN 0 +0 (3.14+1e20)-1e20=0 3.14+(1e20-1e20)=3.14

Representing Instructions int sum(int x, int y) { return x+y; } 00 30 42 Alpha sum 01 80 FA 6B E0 08 81 C3 Sun sum 90 02 00 09 E5 8B 55 89 PC sum 45 0C 03 08 EC 5D C3 For this example, Alpha & Sun use two 4-byte instructions Use differing numbers of instructions in other cases PC uses 7 instructions with lengths 1, 2, and 3 bytes Same for NT and for Linux NT / Linux not fully binary compatible Different machines use totally different instructions and encodings

Machine Words Machine Has “Word Size” Nominal size of integer-valued data Including addresses Most current machines use 32 bits (4 bytes) words Limits addresses to 4GB Becoming too small for memory-intensive applications High-end systems use 64 bits (8 bytes) words Potential address space  1.8 X 1019 bytes Machines support multiple data formats Fractions or multiples of word size Always integral number of bytes

Word-Oriented Memory Organization 32-bit Words 64-bit Words Bytes Addr. 0000 Addresses Specify Byte Locations Address of first byte in word Addresses of successive words differ by 4 (32-bit) or 8 (64-bit) Addr = ?? 0001 0002 0000 Addr = ?? 0003 0004 0000 Addr = ?? 0005 0006 0004 0007 0008 Addr = ?? 0009 0010 0008 Addr = ?? 0011 0008 0012 Addr = ?? 0013 0014 0012 0015

Data Representations Sizes of C Objects (in Bytes) C Data Type Alpha (RIP) Typical 32-bit Intel IA32 unsigned 4 4 4 int 4 4 4 long int 8 4 4 char 1 1 1 short 2 2 2 float 4 4 4 double 8 8 8 long double 8/16† 8 10/12 char * 8 4 4 Or any other pointer (†: Depends on compiler&OS, 128bit FP is done in software)

Byte Ordering How should bytes within multi-byte word be ordered in memory? Conventions Sun’s, Mac’s are “Big Endian” machines Least significant byte has highest address Alphas, PC’s are “Little Endian” machines Least significant byte has lowest address

Byte Ordering Example Big Endian Little Endian Example Least significant byte has highest address Little Endian Least significant byte has lowest address Example Variable x has 4-byte representation 0x01234567 Address given by &x is 0x100 Big Endian 0x100 0x101 0x102 0x103 01 23 45 67 01 23 45 67 Little Endian 0x100 0x101 0x102 0x103 67 45 23 01 67 45 23 01

Boolean algebra Boolean expressions created from: NOT, AND, OR

Digital gate diagram for NOT: Inverts (reverses) a boolean value Truth table for Boolean NOT operator: Digital gate diagram for NOT:

Digital gate diagram for AND: Truth if both are true Truth table for Boolean AND operator: Digital gate diagram for AND:

Digital gate diagram for OR: True if either is true Truth table for Boolean OR operator: Digital gate diagram for OR:

Operator precedence NOT > AND > OR Examples showing the order of operations: Use parentheses to avoid ambiguity

Implementation of gates

Implementation of gates

Implementation of gates Fluid switch (http://www.cs.princeton.edu/introcs/lectures/fluid-computer.swf)

Truth Tables (1 of 3) A Boolean function has one or more Boolean inputs, and returns a single Boolean output. A truth table shows all the inputs and outputs of a Boolean function Example: X  Y

Truth Tables (2 of 3) Example: X  Y

Two-input multiplexer Truth Tables (3 of 3) Example: (Y  S)  (X  S)